基于自适应距离的等距映射

Isometric mapping based on adaptive distance

等距映射是一种非监督的流形学习算法,对于已知先验类别信息的数据,其降维效果不理想。同时,对于新加入的数据点的降维,计算复杂度过高,不适合实时应用。针对这两点不足之处,提出一种推广的自适应等距映射算法,通过引入自适应距离因子并结合推广的等距映射可实现有效降维。ORL人脸数据库的实验表明,该算法在人脸的识别率和实时性上有明显的提高。

Isometric mapping（ISOMAP） is a classic unsupervised algorithm of manifold learning.However,ISOMAP could not work well for the data with the class priori information.Moreover,dimension reduction to the new data point using ISOMAP is computationally too complex to be used in real time.Given the drawbacks of ISOMAP,an adaptive distance generalization of ISOMAP（ADGI） is proposed,in which the adaptive distance factor is introduced to combine with generalization of ISOMAP.The ADGI is effective in the dimension reduction of face features.Experiments on ORL database show the presented algorithm is better both in recognition ratio and in real time for face recognition.